How much is five hours of an underwriter’s time worth?
The obvious answer is, “it depends.”
Is it five hours spent searching a pile of papers to verify a schedule of assets document is correct? Or is it five hours spent developing a new line of business – or building relationships with brokers?
This isn’t an “either/or” question for most underwriting teams. All these jobs need doing—and an underwriter’s time is a fixed cost, no matter the work that’s getting done. But what if non-core functions (like fact-checking documents) could be done quicker?
A lot quicker.
AI acceleration: think fast; act faster
The evolution of commercial AI over the past couple of years has happened shockingly fast, even for an industry intimately familiar with Moore’s Law.
For example, since the much-ballyhooed launch of OpenAI’s ChatGPT 3.5 in November 2022, public Large Language Models (LLMs), including Meta’s Llama 2, Google’s Gemini and TII’s Falcon, have grown exponentially in speed and sophistication.
These public LLMs use billions, even trillions of parameters to create coherent, context-sensitive text to write book-length publications or accurately translate text into hundreds of languages. However, what makes some public LLMs exceptional as chatbots or letting English speakers read novels written in Cantonese does not make them well-suited to underwriting tasks.
For this, an insurance-specific LLM (e.g., InsurGPT™) is the most desirable tool for the job. Trained on the unique rules, taxonomies and tolerances of insurance underwriting, these models interpret a wide range of documents and information to excel at solving an essential underwriting challenge: processing unstructured data.
The value of a quick “no”
“These tools [LLMs] are getting better, faster and more accurate week to week,” said Sean O’Neill, a senior partner at Bain & Co. “That creates an expectation about the art of the possible. In insurance, with the amount of unstructured data, this is tremendously exciting.”
O’Neill addresses “friction” in the underwriting process – around generating schedules of assets and coverages, for example, or the myriad data sources feeding into decisions. AI serves data into those processes in a more actionable format. “Getting it streamlined – and able to deliver the quick ‘no’ [whether to quote] as much as a quick ‘yes’ – is a huge opportunity,” he says
The industry is waking up. “Being able to transform unstructured data into a structured format is transformational,” as Jennifer Krawec, Head of Global Risk Solutions Incubation at Liberty Mutual, puts it. “Plus, insurers are sitting on a mountain of data that’s underutilized. Being able to make use of that data [via rapid automated processing] is game-changing.”
Augmenting Human Expertise
Insurance Document AI isn’t about pointing a tool at a process and then calling it a day. The “first and last miles” of underwriting – at one end, the broker relationship, at the other end, the customer’s decision to purchase a policy – and the creativity to innovate are still fundamentally human activities.
While real-time quoting has become table-stakes in personal lines, commercial insurance underwriting teams need to serve brokers whose expectations for on-demand service are set by Amazon and DoorDash.
“Rapid change forces us to ask: can we build solutions that keep pace with the changes in our industry?” asks John Cottongim, Co-founder & CTO at Roots Automation. “It’s one thing to develop systems based around information retrieval from relatively static sources – legal forms or outputs from your own systems. But on more high-velocity change processes? If it’s external, less predictable data? That’s more challenging.”
Some examples of “quick wins” would be reducing setup time to validate brokers of record and submissions for underwriting, executing straight-through processing of structured data or triaging submissions that best fit carrier appetites.
The acceleration of AI capabilities and the speed at which they handle data will expand the list of use cases. Underwriting teams that understand now how they might use exponentially more capable Insurance Document AI in two years’ time will be the ones reimagining the industry.
Webinar on-demand – “Unlocking Value with AI-Powered Underwriting.”
Want to take steps to ensure these forward-looking underwriters are on your team?
View “Unlocking Value with AI-Powered Underwriting” for more exclusive insights and knowledge on AI for insurance from Sean O’Neill, Jennifer Krawec and John Cottongim.
Click now to discover critical use cases and success stories from insurers using this technology to improve their underwriting results significantly.